Part I of this report noted a number of significant limitations in the sample frame for the Institutional Population Component of NMES. It was noted that there is strong evidence that this caused substantial underrepresentation of smaller community-based residential facilities and their residents in the National Medical Expenditure Survey. The general direction of underrepresentation of small facility residents is clear: "facilities" with 1 or 2 residents were completely eliminated from the study when it became apparent that the sample frame contained only a small proportion of all such facilities nationwide, and there was also considerable underrepresentation of other small facilities.
Corroboration regarding underestimation of small facilities and their residents comes from state reports on the number of facilities that they have under licensure or contract or that they directly operate, and the number of people with mental retardation and related conditions living in them (Lakin, et al., 1989). In addition, the estimation from NMES that the number of small community facilities and residents in 1987 was essentially unchanged from the 1982 NCRF (Lakin, Hill, & Bruininks, 1985), while state institution populations decreased by 25,000 people (most of whom were released into community-based group settings) seems implausible and also suggests underrepresentation of small facilities in the NMES population estimates. As noted earlier, estimates of persons in large facilities were reasonably comparable to expected values.
Given the problems with the estimates in the Institutional Population Component, consideration of ways to adjust this data source to permit more accurate population estimates seems warranted. In the following pages the simplest available method of compensating for the underrepresentation of small facilities is explored. However, it is important to note at the outset that data to assess the precision of any alternative estimates are not available.